INFECTION AND IMMUNITY, Dec. 1979, p. 827-831
Vol. 23, No. 3
0019-9567/79/12-0827/05$02.00/0
Comparison of Two Antigen Detection Techniques in a Primate Model of Haemophilus influenza type b Infection DAVID W. SCHEIFELE,t* ROBERT S. DAUM,4 VASSILIKI PH. SYRIOPOULOU,§ GEORGE R. SIBER,II AND ARNOLD L. SMITH* Division of Infectious Diseases, Children's Hospital Medical Center, and the Department of Pediatrics, Harvard Medical School, Boston, Massachusetts 02115
Received for publication 11 July 1979
Rapid diagnosis of Haemophilus influenza type b meningitis is possible using immunological tests for capsular antigen (polyribophosphate, PRP), such as countercurrent immunoelectrophoresis (CIE) and latex particle agglutination (LPA). We compared two tests in monkeys with evolving, serially quantitated H. influenza type b bacteremia (n = 23) and meningitis (n = 21). In vitro, the LPA test was sensitive to 0.5 ng of PRP/ml of saline, and the CIE test was sensitive to 1.0 ng/ml; in serum, however, CIE detected 5.0 ng of PRP/ml, whereas the sensitivity of LPA was unchanged. LPA detected PRP earlier in the course of bacteremia (mean, 12 h after onset; range, 4 to 36 h) than did CIE (mean, 45 h; range, 4 to 168 h) (P < 0.01). A positive LPA test required 2100 bacteria per ml of blood, whereas CIE required 21,000/ml. PRP accumulated with continuing blood stream infection, aiding detection of low-grade bacteremia. LPA detected antigen in cerebrospinal fluid (CSF) earlier in the course of meningitis and at a lower bacteria density than did CIE. Both methods detected antigen reliably with -1,000 bacteria per ml of CSF. A close correlation existed between CSF concentrations of capsular antigen and bacteria (r = 0.90, P < 0.001). We conclude that the LPA method permits earlier diagnosis of H. influenzae type b infection in part because of its greater sensitivity. b (18). We compared CIE and LPA assays for type b capsular antigen (polyribophosphate, PRP) in monkeys with evolving, serially quantitated H. influenzae type b bacteremia and meningitis. MATERIALS AND METHODS
The recent development of rapid immunological tests for bacterial capsular antigens has added an important new dimension to diagnostic microbiology (7, 14). The most commonly used methods are countercurrent immunoelectrophoresis (CIE) (9, 10, 13, 19) and antibody-coated latex particle agglutination (LPA) (14, 15, 22); both are rapid, simple, and inexpensive. The frequency with which capsular antigen has been detected in clinical trials has varied considerably, depending upon the bacterial species, methods and antiserum used, and site of infection. It has been difficult to describe test sensitivity in functional terms, i.e., the density or duration of infection required to produce a positive test. Such characterization is possible using animal models of infection, such as infant monkeys infected with Haemophilus influenza type t Present address: Infectious Diseases Laboratory, Children's Hospital, Vancouver, British Columbia, Canada V5X1X2. t Present address: Department of Pediatrics, Tulane University Medical School, New Orleans, LA 70112. § Present address: Agia Sophia Children's Hospital, Athen's University, Greece. 1 Present address: Sidney Farber Cancer Institute, Boston, MA 02115. # Present address: Children's Orthopedic Hospital and Medical Center, Seattle, WA 98105.
Infant rhesus monkeys (2 to 30 days old) were inoculated intranasally with 10' colony-forming units (CFU) of H. influenza type b, strain Eag (1, 2). Quantitative blood cultures were obtained 12, 18, and 24 h after inoculation and daily thereafter. Ninety-five percent of bacteremic animals developed meningitis, defined as cerebrospinal fluid (CSF) with pleocytosis and containing H. influenzae Eag. The course of CSF infection was determined by daily quantitative cultures from one or more sites: the lumbar sac, cisterna magna, and lateral ventricle. The methods of quantitative blood and CSF culture have been described (18, 21) and allowed detection of approximately 210 CFU/ml in blood and 2102 CFU/ ml in CSF. Portions of serum and CSF for antigen analysis were stored at -20'C until tested. The CIE and LPA assays have been compared in clinical use (22). The same antiserum was used for each method (burro anti-H. influenzae type b, lot B132, kindly provided by John Robbins, Division of Bacterial Products, Bureau of Biologics, Bethesda, Md.). Purified PRP with a mean molecular weight of
827
828
INFECT. IMMUN.
SCHEIFELE ET AL.
ca. 500,000 (lot 5, provided by Porter Anderson, Rochester, N.Y.) was used to standardize the assays. CIE was performed using 20-tld samples in gels composed of 0.05 M sodium barbital, 0.007 M hydrochloric acid, 0.5% Noble agar (Difco), and 0.5% agarose (pH 8.6). Burro antiserum was diluted 1:4 in phosphatebuffered saline to minimize nonspecific protein precipitation. Electrophoresis was carried out for 30 min at 4VC, using 75 V (30 mA). Results were read immediately. When PRP was diluted in phosphate-buffered saline, the maximal sensitivity was 1.0 ng/ml. In reconstruction experiments with monkey serum (free of anti-PRP antibody) and CSF, the sensitivity was 5.0 and 1.0 ng/ml, respectively. Each batch of samples was tested with positive controls containing 1 and 5 ng/ml PRP in phosphate-buffered saline. The latex agglutination assay was performed as previously described (22). The maximum sensitivity in saline, monkey serum, or CSF, was 0.5 ng/ml after 90 min of agitation. All samples were tested with the same preparation of latex particles; sensitivity did not change throughout the study. Serum samples were also tested with a nonimmune LPA, employing particles coated with normal burro serum, to detect antiglobulins (4). Serum and CSF antigen concentrations were measured using serial dilutions in PBS to the endpoint of each test's sensitivity. When antigen concentration was measured by both CIE and LPA, the same set of dilutions was used and the tests run concurrently. In reconstruction experiments, the values estimated by CIE and LPA differed by twofold or less. Means, linear regressions, correlation coefficients, and significance tests were performed by standard methods (5) using a Wang 360 programmable calculator (Wang Laboratories Inc., Tewksbury, Mass).
RESULTS Twenty-three monkeys developed bacteremia, and 21 had meningitis. A total of 234 samples of serum and 168 samples of CSF were collected at daily intervals during week 1 of infection; each was tested by CIE and cultured quantitatively. Of these, 226 representative samples were selected for LPA testing, comprising 82 CSF and 144 serum samples. Antigen concentration was determined by both methods in the latter samples. No falsely positive tests occurred with CIE or LPA testing of normal pre-inoculation serum (n = 22) or CSF (n = 9). During the week after nasal inoculation with H. influenzae type B, washings of the nasopharynx contained 105 to 107 CFU/ml and 10 to 50 ng/ml of PRP. Ten inoculated animals had a period without detectable bacteremia during which two had 60 h. In 16/18 monkeys in which LPA detected PRP before CIE, the serum antigen concentration was below CIE sensitivity (104 CFU/ ml) and PRP (measured by LPA assay) was
100
IPA0~
F- 2 90 (W
80 70
' |,L
60 -:
1.
I
io
CIE
50 J
40
!j -JW
30
aZX 20 10 L
I,,' 0C
M
6 12 36 60 84 168 DURATION OF BACTEREMIA (hours after onset)
FIG. 1. Relationship between duration of bacteremia and the appearance of detectable antigen in serum, as determined by the LPA and CIE assays in 18 animals.
VOL. 26, 1979
ANTIGEN DETECTION IN HAEMOPHILUS INFECTION
TABLE 1. Antigen detection in serum relative to the density and duration of bacteremia in 18 monkeys Antigen detection (no. positive/no. of animals
tested)
Density of bacteremia
12 ha
(CFU/ml) 102
103 104 >i0O
60 h
36 h
LPA
CIE
LPA
CIE
CIE
3/3 4/4 8/10
1/3 1/4 2/10 1/1
3/3 5/5 8/8 2/2
1/3 4/5 6/8 2/2
1/2 6/7 2/2 5/5
100%
72%
87%
1/1
Detection 89% 28% rate a Duration of bacteremia.
accumulating six times faster than in those without discrepancies. Meningitis developed in 21 animals; comparative CSF data are available from 19. One animal died after 24 h of meningitis. Overall, PRP was detected in the CSF of 19/19 animals using LPA and in 19/20 tested with CIE. The LPA assay detected antigen earlier in the course of meningitis (Fig. 2): in 9/19 animals, LPA was positive before CIE, but in no case was CIE positive before LPA (P < 0.01). Antigen detection was related to CSF bacteria concentrations. At the onset of meningitis, bacteria density ranged from 102 to 106 CFU/ml. LPA detected antigen in the CSF of all 12 animals with 2103 CFU/ml and in 5/7 (71%) with 102 CFU/ml. CIE detected antigen in 6/7 with 2104 CFU/ml, but in only 4/13 (30%) with fewer bacteria. Twenty-four hours later, LPA detected PRP in all animals (n = 19) whose CSF contained 102 to 106 CFU/ml; CIE revealed antigen in 15/15 animals infected with 2103 CFU/ml and in 1/3 with 102 CFU/ml. The one falsenegative result with CIE occurred in a monkey with 102 CFU/ml of CSF for 4 days. The lower detection rate with CIE was a function of low CSF PRP concentrations; all 10 failures had
Vol. 23, No. 3
0019-9567/79/12-0827/05$02.00/0
Comparison of Two Antigen Detection Techniques in a Primate Model of Haemophilus influenza type b Infection DAVID W. SCHEIFELE,t* ROBERT S. DAUM,4 VASSILIKI PH. SYRIOPOULOU,§ GEORGE R. SIBER,II AND ARNOLD L. SMITH* Division of Infectious Diseases, Children's Hospital Medical Center, and the Department of Pediatrics, Harvard Medical School, Boston, Massachusetts 02115
Received for publication 11 July 1979
Rapid diagnosis of Haemophilus influenza type b meningitis is possible using immunological tests for capsular antigen (polyribophosphate, PRP), such as countercurrent immunoelectrophoresis (CIE) and latex particle agglutination (LPA). We compared two tests in monkeys with evolving, serially quantitated H. influenza type b bacteremia (n = 23) and meningitis (n = 21). In vitro, the LPA test was sensitive to 0.5 ng of PRP/ml of saline, and the CIE test was sensitive to 1.0 ng/ml; in serum, however, CIE detected 5.0 ng of PRP/ml, whereas the sensitivity of LPA was unchanged. LPA detected PRP earlier in the course of bacteremia (mean, 12 h after onset; range, 4 to 36 h) than did CIE (mean, 45 h; range, 4 to 168 h) (P < 0.01). A positive LPA test required 2100 bacteria per ml of blood, whereas CIE required 21,000/ml. PRP accumulated with continuing blood stream infection, aiding detection of low-grade bacteremia. LPA detected antigen in cerebrospinal fluid (CSF) earlier in the course of meningitis and at a lower bacteria density than did CIE. Both methods detected antigen reliably with -1,000 bacteria per ml of CSF. A close correlation existed between CSF concentrations of capsular antigen and bacteria (r = 0.90, P < 0.001). We conclude that the LPA method permits earlier diagnosis of H. influenzae type b infection in part because of its greater sensitivity. b (18). We compared CIE and LPA assays for type b capsular antigen (polyribophosphate, PRP) in monkeys with evolving, serially quantitated H. influenzae type b bacteremia and meningitis. MATERIALS AND METHODS
The recent development of rapid immunological tests for bacterial capsular antigens has added an important new dimension to diagnostic microbiology (7, 14). The most commonly used methods are countercurrent immunoelectrophoresis (CIE) (9, 10, 13, 19) and antibody-coated latex particle agglutination (LPA) (14, 15, 22); both are rapid, simple, and inexpensive. The frequency with which capsular antigen has been detected in clinical trials has varied considerably, depending upon the bacterial species, methods and antiserum used, and site of infection. It has been difficult to describe test sensitivity in functional terms, i.e., the density or duration of infection required to produce a positive test. Such characterization is possible using animal models of infection, such as infant monkeys infected with Haemophilus influenza type t Present address: Infectious Diseases Laboratory, Children's Hospital, Vancouver, British Columbia, Canada V5X1X2. t Present address: Department of Pediatrics, Tulane University Medical School, New Orleans, LA 70112. § Present address: Agia Sophia Children's Hospital, Athen's University, Greece. 1 Present address: Sidney Farber Cancer Institute, Boston, MA 02115. # Present address: Children's Orthopedic Hospital and Medical Center, Seattle, WA 98105.
Infant rhesus monkeys (2 to 30 days old) were inoculated intranasally with 10' colony-forming units (CFU) of H. influenza type b, strain Eag (1, 2). Quantitative blood cultures were obtained 12, 18, and 24 h after inoculation and daily thereafter. Ninety-five percent of bacteremic animals developed meningitis, defined as cerebrospinal fluid (CSF) with pleocytosis and containing H. influenzae Eag. The course of CSF infection was determined by daily quantitative cultures from one or more sites: the lumbar sac, cisterna magna, and lateral ventricle. The methods of quantitative blood and CSF culture have been described (18, 21) and allowed detection of approximately 210 CFU/ml in blood and 2102 CFU/ ml in CSF. Portions of serum and CSF for antigen analysis were stored at -20'C until tested. The CIE and LPA assays have been compared in clinical use (22). The same antiserum was used for each method (burro anti-H. influenzae type b, lot B132, kindly provided by John Robbins, Division of Bacterial Products, Bureau of Biologics, Bethesda, Md.). Purified PRP with a mean molecular weight of
827
828
INFECT. IMMUN.
SCHEIFELE ET AL.
ca. 500,000 (lot 5, provided by Porter Anderson, Rochester, N.Y.) was used to standardize the assays. CIE was performed using 20-tld samples in gels composed of 0.05 M sodium barbital, 0.007 M hydrochloric acid, 0.5% Noble agar (Difco), and 0.5% agarose (pH 8.6). Burro antiserum was diluted 1:4 in phosphatebuffered saline to minimize nonspecific protein precipitation. Electrophoresis was carried out for 30 min at 4VC, using 75 V (30 mA). Results were read immediately. When PRP was diluted in phosphate-buffered saline, the maximal sensitivity was 1.0 ng/ml. In reconstruction experiments with monkey serum (free of anti-PRP antibody) and CSF, the sensitivity was 5.0 and 1.0 ng/ml, respectively. Each batch of samples was tested with positive controls containing 1 and 5 ng/ml PRP in phosphate-buffered saline. The latex agglutination assay was performed as previously described (22). The maximum sensitivity in saline, monkey serum, or CSF, was 0.5 ng/ml after 90 min of agitation. All samples were tested with the same preparation of latex particles; sensitivity did not change throughout the study. Serum samples were also tested with a nonimmune LPA, employing particles coated with normal burro serum, to detect antiglobulins (4). Serum and CSF antigen concentrations were measured using serial dilutions in PBS to the endpoint of each test's sensitivity. When antigen concentration was measured by both CIE and LPA, the same set of dilutions was used and the tests run concurrently. In reconstruction experiments, the values estimated by CIE and LPA differed by twofold or less. Means, linear regressions, correlation coefficients, and significance tests were performed by standard methods (5) using a Wang 360 programmable calculator (Wang Laboratories Inc., Tewksbury, Mass).
RESULTS Twenty-three monkeys developed bacteremia, and 21 had meningitis. A total of 234 samples of serum and 168 samples of CSF were collected at daily intervals during week 1 of infection; each was tested by CIE and cultured quantitatively. Of these, 226 representative samples were selected for LPA testing, comprising 82 CSF and 144 serum samples. Antigen concentration was determined by both methods in the latter samples. No falsely positive tests occurred with CIE or LPA testing of normal pre-inoculation serum (n = 22) or CSF (n = 9). During the week after nasal inoculation with H. influenzae type B, washings of the nasopharynx contained 105 to 107 CFU/ml and 10 to 50 ng/ml of PRP. Ten inoculated animals had a period without detectable bacteremia during which two had 60 h. In 16/18 monkeys in which LPA detected PRP before CIE, the serum antigen concentration was below CIE sensitivity (104 CFU/ ml) and PRP (measured by LPA assay) was
100
IPA0~
F- 2 90 (W
80 70
' |,L
60 -:
1.
I
io
CIE
50 J
40
!j -JW
30
aZX 20 10 L
I,,' 0C
M
6 12 36 60 84 168 DURATION OF BACTEREMIA (hours after onset)
FIG. 1. Relationship between duration of bacteremia and the appearance of detectable antigen in serum, as determined by the LPA and CIE assays in 18 animals.
VOL. 26, 1979
ANTIGEN DETECTION IN HAEMOPHILUS INFECTION
TABLE 1. Antigen detection in serum relative to the density and duration of bacteremia in 18 monkeys Antigen detection (no. positive/no. of animals
tested)
Density of bacteremia
12 ha
(CFU/ml) 102
103 104 >i0O
60 h
36 h
LPA
CIE
LPA
CIE
CIE
3/3 4/4 8/10
1/3 1/4 2/10 1/1
3/3 5/5 8/8 2/2
1/3 4/5 6/8 2/2
1/2 6/7 2/2 5/5
100%
72%
87%
1/1
Detection 89% 28% rate a Duration of bacteremia.
accumulating six times faster than in those without discrepancies. Meningitis developed in 21 animals; comparative CSF data are available from 19. One animal died after 24 h of meningitis. Overall, PRP was detected in the CSF of 19/19 animals using LPA and in 19/20 tested with CIE. The LPA assay detected antigen earlier in the course of meningitis (Fig. 2): in 9/19 animals, LPA was positive before CIE, but in no case was CIE positive before LPA (P < 0.01). Antigen detection was related to CSF bacteria concentrations. At the onset of meningitis, bacteria density ranged from 102 to 106 CFU/ml. LPA detected antigen in the CSF of all 12 animals with 2103 CFU/ml and in 5/7 (71%) with 102 CFU/ml. CIE detected antigen in 6/7 with 2104 CFU/ml, but in only 4/13 (30%) with fewer bacteria. Twenty-four hours later, LPA detected PRP in all animals (n = 19) whose CSF contained 102 to 106 CFU/ml; CIE revealed antigen in 15/15 animals infected with 2103 CFU/ml and in 1/3 with 102 CFU/ml. The one falsenegative result with CIE occurred in a monkey with 102 CFU/ml of CSF for 4 days. The lower detection rate with CIE was a function of low CSF PRP concentrations; all 10 failures had
